Sovereign debt crises

Latin America: the lost decade: 1980s

Mexico, 1985

Argentina, 1992

Ecuador, 2009

Asia, 1982, 1997

Africa, Two lost decades, culminating in HIPC initiative

2

40 years of sovereign debt crises

o A country is said to experience a debt crisis in a year if one of the following conditions holds: The sum of its interest and principal arrears on

long-term debt outstanding to all creditors is larger than 5% of the total debt outstanding.

The country receives debt relief from the Paris Club (w/o HIPC initiative exits)

The country receives substantial balance-of-payment support from the IMF in the form of StandBy Arrangements and Extended Fund Facility. We define exceptional support by the IMF the event where a country actually uses more than 50% of its quota in one year (KN just looked at events where the IMF extended resources to the country in excess of 50% of its quota, regardless of the fact that it actually used it).

o Data on 126 countries from 1970 to 2007.

3

Debt crises: a simple but misleading statistic

Unconditional probability for a country to experience a debt crisis in any given year : 37%

Very high number! How can we explain such a high

default probability? Serial defaulters? Global crises?

4

The serial defaulters hypothesis (1)

010

20

30

40

Nu

mbe

r o

f D

efa

ults

0 50 100 150Country

Number of Defaults By Country

5

The serial defaulters hypothesis (2)

On average a given country in our sample is considered to be in debt distress one third of the time (11,6 years out of 38 years).

When we exclude the countries which spend more than 20 years out of 38 in a debt crisis, i.e. the “serial defaulters”, the mean annual probability of a debt crisis only goes to 24%.

6

The global crisis hypothesis (1)

020

40

60

Nu

mbe

r o

f de

faults

1970 1980 1990 2000 2010Year

Number of Defaults By Year

7

The global crisis hypothesis (2)

On a given year, one third of the countries (38,5 out of 126) are in debt distress on average.

When we exclude the years where more than 52 countries are in debt crisis (the 4th quartile), the mean annual probability of a debt crisis does not change that much: 32%.

8

Market access?

default probability

market access countries

27%

non-market access countries

43%

9

US business cycle?

default probability

years where there is a recession in the US

(NBER)

33%

years where there is no recession in the US

(NBER)

38%

10

A new definition of a debt crisis

Debt distress episodes: years where a country experiences a debt crisis as previously defined, iif this crisis is preceded by three years without crisis.

Similarly, normal times: years without any crisis preceded by three years without any crisis.

Avoids counting several consecutive years of crisis as different debt distress events.

What we thought was a high unconditional default probability is in fact the sign that debt crises are protracted in most countries. New unconditional default probability: 7%.

11

Serial Defaulters revisited

01

23

4N

um

be

r o

f D

efa

ults

0 50 100 150Country

Number of Defaults By Country

Outliers: The Gambia (3 times), Ghana (3), Grenada (3), Kenya (4), Turkey (3) and Uruguay (3).

12

Global crisis revisited

05

10

15

Nu

mbe

r o

f de

faults

1970 1980 1990 2000 2010Year

Number of Defaults By Year

13

Non-Market Access Countries experience longer crises

0.0

2.0

4.0

6

0 1

Fre

qu

en

cy o

f D

efa

ults

Graphs by marketaccess

Default Frequency By Market Access

14

Debt crises, Currency crises and Banking crises

Definitions of Laeven and Valencia (2008) to identify both currency crises and banking crises

36 simultaneous sovereign and currency crises, 19 simultaneous sovereign and banking crises, 40 simultaneous currency and banking crises and 10 triple crises.

This tends to show that debt crises are not so often correlated with other kinds of crises but are rather crises of their own.

15

The macroeconomic determinants of debt crises

o Consider the following model: P (yct=1) = G (β’Xct)

● where yct is a dummy variable equal to 1 when country c experienced a debt crisis at time t and 0 otherwise. Xct is a vector of explanatory variables, β is the vector of estimated coefficients and G is the cumulative distribution function of the logistic distribution (logit estimation).

● Independent variables: log(real GDP per capita), log(debt/GDP), total debt service/exports, CPIA all lagged for two years and the spread between the yield of corporate bonds in the US rated Baa by Moody’s and the yield of 10-year US Treasury bonds (proxy for world financial shock).

16

Resultsdebt

distress=1

log (debt/GDP)t-2 0,523***

total debt service/exportst-2

3,943***

real gdp per capitat-2 -0,379***

cpiat-2 -0,577***

Baa US corporates spread

0,611***

intercept 0,718

N 1159

Pseudo R² 0,0987

Prob>Chi2 0,0000

*** denotes 1% significance, ** denotes 5% significance, *

denotes 10% significance

17

Quintiles of risk (1)

default probability

A p < 2,5%

B 2,5%< p <4,4%

C 4,4%< p <7,1%

D 7,1%< p <11,3%

E 11,3%< p

18

Quintiles of risk (2)

A B C D E TOTAL

Default

events

(number) 4 10 16 16 42 88

Non-

Default

events

(number)) 227 222 216 216 190 1071

TOTAL 231 232 232 232 232

19

Dynamics in risk classification

1 2 3 4 5

At-3 Bt-3 Ct-3 Dt-3 Et-3TOTAL

1

Defaulters

At 0 0 0 2 0 2

2

Defaulters

Bt 4 0 1 1 0 6

3

Defaulters

Ct 0 1 2 3 1 7

4

Defaulters

Dt 1 4 2 4 1 12

5

Defaulters

Et 0 5 11 10 22 48

TOTAL 5 10 16 20 24  75

20

The sources of risk (1)

We want to focus on the respective weight carried by each factor identified by our baseline regression in the total default probability. Four factors stood out in our estimation of the likelihood of a debt distress episode:

Debt and Debt service GDP per capita Governance quality (CPIA) World shock (Baa-Treasury Bonds spread)

Let’s call z the risk index corresponding to the linear combination of these risk factors weighted by the coefficients of the baseline regression. The default probabilities we estimate are simply G (z) where G is the cumulative distribution function of the logistic distribution.

The z factor is additive: we can directly weight the influence of such or such terms on the probability of default.

21

The sources of risk (2)

In order to measure the influence of each variable on the default probability, we computed the average value of each variable (weighted by its coefficient in the regression) for each risk category A, B, C, D and E.

A B C D E

Governance

quality -2,3 -2,1 -1,9 -1,8 -1,6

I ntercept 0,7 0,7 0,7 0,7 0,7

z - 4,1 - 3,4 - 2,8 - 2,4 - 1,5

1,3 1,4

World

shock 1,2 1,3 1,3

0,05 0,3 0,8

GDP per

capita -3,2 -3,1 -3 -2,9 -2,8

Debt+Debt

service -0,5 -0,2

22

Risk Probabilities (1)

We take the average values of each variable for class E countries as a numeraire and we compute the role of each factor in explaining the overall risk.

We measure by how much the risk in decreased when switching the average of each of the five variables in E to the average in A, B, C and D respectively. Compared to the global effect z, we can measure the contribution of each factor to total risk.

In the following table we compute for each cell P(xi+vE) where xi+vi =zi. All x, v and z are averaged over i=A, B...E

23

Risk Probabilities (2)

i=A i=B i=C i=D i=E

x=

Governance 0,1 0,12 0,14 0,15 0,18

z 0,02 0,03 0,06 0,08 0,18

x=

Debt+Debt

service 0,06 0,08 0,1

0,14 0,15

0,12 0,18

0,17 0,18

x= World

shock 0,15 0,17 0,17 0,17 0,18

x= GDP per

capita 0,13

24

today

o In the last crisis, 2008-2010, sovereign debt crises was much less severe than one could have expected

o Most vulnerable countries were either non market countries, or already in crisis…

o The big surprise was the sovereign debt crisis in EUROPE

25

Greek crisis

26

Chronicle (2009-2010)

On the 3 september 2009, PM Costas Caramanlis calls new elections

4 october 2009, elections won by George Papandreou

21 october 2009, budget shoot from 3.7% expected to 12.5%

27

The spiral

0

200

400

600

800

1000

1200

1400

1-se

pt-2

009

15-s

ept-2

009

29-s

ept-2

009

13-o

ct-2

009

27-o

ct-2

009

10-n

ov-2

009

24-n

ov-2

009

8-dé

c-20

09

22-d

éc-2

009

5-jan

v-20

10

19-ja

nv-2

010

2-fé

vr-2

010

16-fé

vr-2

010

2-m

ars-

2010

16-m

ars-

2010

30-m

ars-

2010

13-a

vr-2

010

27-a

vr-2

010

11-m

ai-20

10

25-m

ai-20

10

8-ju

in-2

010

22-ju

in-2

010

6-ju

il-20

10

BPs

October 4 - Papandreou's Panhellenic Socialist Movement (PASOK) party wins power. At this stage, the government expects the 2009 deficit to reach 6% of GDP

December 16 - S&P cuts Greece’s rating to BBB+ from A-

December 22 - Moody’s cuts Greek debt to A2 from A1

January 14 - Greece unveils a stability program w/ budget gap cut to 2.8% of GDP in 2012 from 12.7 in 2009

February 3 - The EU Commission says it backs Greece’s plan to reduce its budget deficit below 3% of GDP by 2012

March 5 - New package of public sector pay cuts and tax increases

March 25 - Euro zone leaders agree to create a joint financial safety net for Greece with the IMF

April 11 - Euro zone finance ministers approve a €30 bn emergency aid mechanism for Greece

April 27 - S&P downgrades Greece’s credit rating to junk status

May 2 - Greece seals deal with EU and IMF opening door to 110 bn euro bailout in exchange for extra budget cuts of €30 bn over 3 years

May 18 - Greece receives a € 14.5 bn loan from the EU and can now repay its immediate debt

December 8 - Fitch Ratings cuts Greek debt to BBB+ w/ a negative outlook

November 5 – The new government pledges to save Greece from banruptcy by cutting the budget deficit

June 14 - Moody’s cuts Greece’s credit rating 4 notches to Ba1 (or junk status) over risks to an EU / IMF package

February 2 - Government extends a public sector wage freeze to those making below €2,000 a month for 2010

28

The European response

30 bls euros on 11 april 2010

110 bls euros on May, along with a program aiming at a 2.6% deficit in 2014

On July 21st, 2011, another 109 bl package is voted, cum a market participation of 106 bn, involving a 21% haircut.

29

Former European episodes

Ireland 87-89 Debt to GDP from de 107% à 92.5%

Structural deficit from –7% to –2.2%

Dévaluation beforehand

I RE L AN D

1983 1984 1985 1986 1987 1988 1989 1990 1991 1992

Real GDP Growth -0.7 2.7 2.6 0.7 3.2 3.1 6 8.3 1.9 3.3

General government structural balance in percent of l GDP

-10.1 -8.3 -9.5 -8.7 -7 -3.4 -2.2 -4.5 -2.7 -1.6

Primary Balance / GDP -2.7 -1.5 0.8 2.8 4.2 2.1 2.6 2.8

Public Debt Stock / GDP 97.2 102.8 104.7 104.4 107 102.3 92.5 86.9 84.6 82.6

Exchange Rate (to US$) 1.134996482 0.991502821 1.243502698 1.399501777 1.675490919 1.507499812 1.556299121 1.775505131 1.749811895 1.629407547

30

Former European episodes (2)

Sweden 94-98 Budget went from a structural deficit of –11.8% in 1994 to a

surplus of 0.9% in 1998

Devaluation preceded the crisis

SWE DE N

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Real GDP Growth -1.1 -1.4 -2.2 4.1 3.7 1.1 2.1 3.6 4.5 3.6

General government structural balance in percent of potential GDP -3.7 -5.7 -13.1 -11.8 -8.8 -4.9 -3.7 0.6 1.8 4.6

Deficit / GDP -8.9 -11.2 -9.1 -7.3 -3.3 -1.6 0.9 0.8 3.6

Primary Balance / GDP -4.0 -5.0 -4.6 -3.7 0.5 2.6 2.8 2.8 4.4

Interest Charge on Public Debt / GDP 0.4 0.9 1.7 2.4 2.8 3.0 2.6 2.5 2.1

Public Debt Stock / GDP 73.4 78.2 82.5 81.1 84.4 83.0 82.0 73.2 64.3

Exchange Rate (to US$) 5.5295 7.043 8.3035 7.4615 6.6582 6.871 7.877 8.061 8.525 9.535

31

Former European episodes (3)

IRELAND SUÈDE

32

Previous Greek adjustment1992-2002

-8,0

-6,0

-4,0

-2,0

0,0

2,0

4,0

6,0

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

%

General Government cyclically adjusted primary balances

33

Theoretical considerations

What is the nature of Greek crisis?

Solvency crisis?

Self-fulfilling debt crisis? Debt service jumps from i*D/Q to (i+p)D/Q

What should Europe do? Set up a lender of first resort facility?

34

Thoughts on ratings agencies

They play a critical role, explaining government fears

What is their role? Do they follow or anticipate market sentiments?

Two conflicting views Procyclical: Kaminsky et Schmukler (2001); Ferri et Stiglitz

(1999), Reisen et Malzan (1998)

‘Neutral’, often late: Reinhart (2002), Gaillard (2009)

Spreads et notations: l’Uruguay 2002

Spreads and ratings: Brazil, 1998-99

37

Spreads and ratings: Greece

Spreads before and after « watch negative »

Spreads before and after a positive outlook annoucement

Policy agenda

European ratings agency

Role of ECB

Lender of first resort (ESF)

Structural deficits targets

Cooperation monetary and fiscal policies

….

40